A Structured Evidence Selection framework (SESAD) that reformulates anomaly detection as a structured reasoning process over clip-level visual evidence and performs context-conditioned selection under scene and action constraints, thereby alleviating semantic entanglement under weak supervision.
Abstract
Weakly supervised video anomaly detection relies solely on video-level labels for training, making it difficult to accurately localize anomalous events in complex scenes. In real-world videos, anomalous behaviors exhibit large variations in appearance and temporal duration, while scene appearance and action dynamics are often tightly entangled. Consequently, existing models tend to rely on scene-related statistical cues rather than true behavioral deviations, resulting in unstable detection performance. To address this challenge, we propose a Structured Evidence Selection framework (SESAD) that reformulates anomaly detection as a structured reasoning process over clip-level visual evidence. Instead of directly mapping aggregated features to anomaly scores, SESAD reorganizes clip representations into semantically structured candidate evidence and performs context-conditioned selection under scene and action constraints. This mechanism adaptively emphasizes anomaly-relevant semantics while suppressing scene interference, thereby alleviating semantic entanglement under weak supervision. Furthermore, we introduce a lightweight geometric discrimination module that constructs a dual-prototype structure in the embedding space, enabling anomaly decisions through relative geometric relations. Extensive experiments on UBnormal, ShanghaiTech, and UCF-Crime show that SESAD achieves 67.92, 97.99, and 88.46 AUC, respectively, while maintaining high computational efficiency and overall consistently stable anomaly discrimination.
Experiments on UCF-Crime and UBnormal show that CSI-VAD consistently improves over the direct holistic baseline and achieves competitive performance against existing methods, showing the advantage of structured context decomposition for training-free video anomaly detection.
Dongjun Kim, Changjae Oh, Andrea Cavallaro et al.· arXiv.org· 0 citations
Detecting anomalous events in surveillance videos is a critical yet challenging task due to the rarity, diversity, and unpredictable nature of abnormal activities. Existing methods often rely on fully supervised annotations or weakly supervised multiple instance learning frameworks that require labeled anomalous videos and complex training strategies. In this work, we propose AAAD (Action-Aware Anomaly Detection), a framework that models normal human behavior using learned action embeddings and detects anomalies as semantic deviations. The proposed method first segments videos into fixed-length clips and extracts compact action embeddings using a frozen ResNet18 backbone. An autoencoder is then trained exclusively on embeddings derived from normal clips, enabling the model to learn the distribution of normal actions without requiring anomaly labels. During inference, anomalies are detected based on reconstruction error in the embedding space, allowing temporal localization of abnormal events at the clip level. Experiments conducted on the UCF-Crime dataset demonstrate that the proposed approach effectively distinguishes normal and abnormal activities, achieving 69.0% AUC with a separation ratio of 1.25x. Our method outperforms unsupervised baselines including k-NN (56.4%), One-Class SVM (60.5%), Isolation Forest (54.5%), and Clustering (57.0%). Qualitative and quantitative results confirm that modeling normal action semantics provides a robust and scalable solution for real-world surveillance anomaly detection.
Mahmoud Elnady, H. E. Abdelmunim· Discover Computing· 0 citations
Despite growing interest in weakly supervised video anomaly detection (WSVAD), current methods struggle to bridge the gap between coarse temporal supervision and fine-grained spatial reasoning. A key obstacle is the tendency of temporal detectors to latch onto background and scene-level cues rather than truly discriminative anomaly evidence. This background bias raises ethical concerns: models may inadvertently associate anomalies with societal or environmental context rather than authentic crime-related cues. Without spatial grounding, such biases remain hidden and unauditable. To address this, we propose SST-WSVADL, a sparse spatio-temporal framework that bridges temporal anomaly detection with fine-grained spatial localization. Rather than processing all spatial regions indiscriminately, SST-WSVADL progressively focuses on the most anomaly-relevant spatio-temporal regions through dynamic sparsification, naturally suppressing background dominant content while preserving discriminative evidence. The temporal and spatial branches are coupled end-to-end via motion-aware regularization that guides sparsification toward dynamically informative regions, without relying on external detectors or vision-language prompts. We publicly release frame-level spatial annotations and a method-agnostic evaluation protocol for three public datasets: UCF-Crime, XD-Violence, and MSAD. These resources enable the community to audit spatial biases in WSVAD predictions, supporting progress toward more ethical and accountable anomaly detection. Experiments demonstrate that SST-WSVADL is competitive with prior methods across benchmarks while enabling localization and patch-level auditability of scene bias, providing a reproducible foundation for interpretability-oriented evaluation of WSVAD models.
S. Abdulaziz, Abdulrahman Al-Abri, Giacomo D'Amicantonio et al.· 1 citation
As the scale of video surveillance data outpaces manual annotation capacities, weakly supervised video anomaly detection (WSVAD) has emerged as a critical research frontier. Most existing approaches formulate WSVAD within a Multiple Instance Learning (MIL) framework that relies on rigid, hand-crafted temporal priors to supervise anomaly scoring. However, such formulations exhibit limited adaptability to the wide variation in anomaly durations and temporal dynamics observed in real-world videos, often leading to unstable or unreliable snippet-level predictions. To address this limitation, we propose an adaptive temporal modeling framework for WSVAD that explicitly accounts for variations in video dynamics across multiple temporal granularities. First, we introduce a Temporal Refinement Module (TRM) that leverages dynamic positional encoding and a learnable class token to model long-range temporal dependencies while distilling a stable global video-level representation. Second, to capture anomalous events with varying frequency and duration, we develop an adaptive Event Segmentation Module (ESM) that identifies event boundaries through temporal discontinuity analysis and aggregates snippet features into discriminative event-level representations. Finally, for snippet-level and event-level predictions, we propose an adaptive similarity-based fusion strategy that dynamically integrates anomaly scores into video-level predictions, replacing fixed top-k aggregation heuristics with global semantic relevance. Extensive experiments on two benchmarks demonstrate that the proposed framework consistently outperforms state-of-the-art methods.
Video Anomaly Detection (VAD) is a crucial computer vision task for
security monitoring and public safety. Unsupervised VAD is more suitable for
real-world scenarios with rare unknown anomalies, but existing LLM-based
methods suffer from limited temporal modeling, inconsistent video understand
ing and inaccurate fine-grained localization, leading to biased anomaly scoring.
To solve these problems, we propose a novel unsupervised VAD framework fus
ing graph attention propagation and multimodal semantic information: first, fuse
video semantic and motion features to construct a dynamic spatiotemporal graph,
and refine node features via graph attention propagation with orthogonal con
straints; then, split videos into semantically coherent event units by a statistical
boundary detection module; finally, guide MLLMs to generate event semantic
descriptions and initial anomaly scores through a hierarchical prompting strategy,
and refine the scores via video-text semantic alignment to obtain accurate frame
level scores. Evaluated on UCF-Crime and XD-Violence datasets with frame
level AUC, the proposed framework achieves state-of-the-art performance under
unsupervised and zero-shot settings, significantly outperforming existing LLM
based VAD methods and even several weakly supervised approaches, which fully
verifies its effectiveness and robustness.
Qinghao Kong· Poster Volume 0008 The 2026...· 0 citations
TAU-Bench is introduced, a track-centric benchmark for jointly evaluating anomaly instance tracking and fine-grained anomaly understanding, and shows that models producing plausible anomaly interpretations may still fail to localize and track the correct instance reliably, revealing a persistent gap between semantic reasoning and visual grounding.
Kepeng Yang, Dong-Xuan Liu, Rongxin Gao et al.· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
